Rate limits
Keep machine access predictable and friendly to the app.
Current guidance
Avoid tight polling loops and bulk writes without pauses. MCP/API tokens should be used for user-initiated work and bounded automation.
Failure handling
Clients should handle 401 and 403 responses by checking token status, and should back off on 429 or transient server errors when formal limits are added.
Future limits
Published per-token and per-customer limits should be added here when they are enforced in the backend.
// docs maintenance
Keep docs updated with product, API, UI, and workflow changes. If behavior changes, the matching docs page should change in the same work.